SUBJECT: Re : interpolation OK , the problem is this : I 've already got log ( &CHAR ( &CHAR ) ) and log ( &CHAR ( &CHAR ) ) , because I 'm calculating probabilities in log space to avoid underflow . Thus , I have to find the equivalent of interpolating in the normal probability space . &NAME is that normal interpolation requires addition which does n't have an equivalent in the log space . Thus , the only thing I can think of is to scale and convert back to normal probs , but this still might cause problems with under / overflow if the probabilities are markedly different . Any further ideas , I 'd be grateful . Thanks &NAME , &NAME On &NAME , &NUM &NAME &NUM , &NAME &NAME wrote : <QUOTE> Hi , I 'm not sure but I suspect it would be significantly more difficult in the log space since we &CHAR trying to find a components of a log addition log ( &CHAR ( &CHAR ) &CHAR ( &CHAR ( &CHAR ) &CHAR ( &CHAR ) ) in terms of log ( &CHAR ) and Log ( &CHAR ) maybe sticky . Will check with the other experts here &SMILEY - no one around now . Do you not want to convert them back and forth ? - &NAME On &NAME , &NUM &NAME &NUM , &NAME &NAME wrote : <QUOTE> &NAME , How would you interpolate log-probs ? i.e. &NAME ( &CHAR ) &CHAR ( &CHAR ) + &NAME ( &CHAR ) &CHAR ( &CHAR ) - in standard probs ? ? - in log probs Any ideas ? &CHAR <END_QUOTE> <END_QUOTE>